Retrieval of Cloud Optical Thickness from Sky-View Camera Images using a Deep Convolutional Neural Network based on Three-Dimensional Radiative Transfer

Masuda, Ryosuke; Iwabuchi, Hironobu; Schmidt, S.; Damiani, Alessandro; Kudo, Rei

doi:10.3390/rs11171962

Cited by 26 publications

(20 citation statements)

References 46 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Based on digital sky images recorded by a previously calibrated sky camera system, we estimated the COD spatial distribution using a deep-learning algorithm based on 3D radiative transfer (Masuda et al 2019). The angular distribution of visible transmitted radiance training data was generated by simulations using Monte Carlo Atmospheric Radiative Transfer Simulator (MCARaTS), a 3D radiative transfer model (Iwabuchi 2006), assuming various 3D cloud fields obtained from highresolution cloud dynamical simulations.…”

Section: Resultsmentioning

confidence: 99%

An Intensive Campaign-Based Intercomparison of Cloud Optical Depth from Ground and Satellite Instruments under Overcast Conditions

Damiani

Irie

Takamura

et al. 2019

SOLA

Self Cite

View full text Add to dashboard Cite

We used observations recorded at Chiba University in November 2018 to examine the variability in cloud optical depth (COD) under overcast conditions. First, we conducted a careful evaluation of four COD datasets retrieved from three types of surface observations: i) zenith radiance recorded by two sky radiometers; ii) solar radiation data collected by a pyranometer; and iii) spatial distribution of radiance recorded using a sky camera system. Although the COD retrieved from the pyranometer (camera) slightly (moderately) overestimated the COD from zenith radiance, we found a satisfactory correlation among all surface estimates. This result suggests the efficacy of both pyranometer-and camera-based approaches and supports their broader use when dedicated cloud observations are not available. We then assessed satellite-based COD estimates retrieved from the recently launch ed Advanced Himawari Imager (AHI) aboard Himawari-8 (H-8) and Second-generation Global Imager (SGLI) on the Global Change Observation Mission for Climate (GCOM-C). Overall, we found good agreement between ground and satellite estimates; their correlation and root mean square error were virtually equivalent to values reported for co-located surface-based instruments. Nevertheless, the AHI-based COD was found to be slightly positively biased with respect to surface datasets.

show abstract

Section: Resultsmentioning

confidence: 99%

An Intensive Campaign-Based Intercomparison of Cloud Optical Depth from Ground and Satellite Instruments under Overcast Conditions

Damiani

Irie

Takamura

et al. 2019

SOLA

Self Cite

View full text Add to dashboard Cite

show abstract

“…Cloud field properties are derived from LES and SSI from 3‐D radiative transfer, providing known inputs and outputs and the need to learn the complex relationship between them. This problem is an ideal candidate for machine learning, which is increasingly being adopted in the atmospheric sciences (e.g., Glassmeier et al, 2019; Masuda et al, 2019; McGovern et al, 2017; O'Gorman & Dwyer, 2018; Thampi et al, 2017). One of the common criticisms of machine learning approaches is that they are used as a “black box” providing little physical interpretability.…”

Section: Introductionmentioning

confidence: 99%

On the Relationship Between Shallow Cumulus Cloud Field Properties and Surface Solar Irradiance

Gristey

Feingold

Glenn³

et al. 2020

Geophysical Research Letters

Self Cite

View full text Add to dashboard Cite

Shallow cumulus clouds exhibit highly three-dimensional (3-D) spatial structure leading to complex variability in the surface solar irradiance (SSI) beneath. This variability is captured by the typically bimodal shape of the SSI probability density function (PDF). Using large eddy simulation to generate well-resolved cloud fields and Monte Carlo 3-D radiative transfer to reproduce realistic associated SSI PDFs, we seek direct relationships between the cloud field properties and the SSI PDF shape. Applying both random forest and artificial neural network algorithms, we find variations in the two modes of the SSI PDF are well predicted by just a handful of cloud field properties. The two algorithms utilize cloud properties similarly, with indistinguishable performance despite their different architectures. These results offer a marked improvement in realism relative to one-dimensional radiative transfer while bypassing computationally expensive 3-D radiative transfer, with immediate application to renewable energy assessments, and potential for several other geophysical applications. Plain Language Summary When sunlight is intercepted by broken clouds, it is not straightforward to calculate how much of the sunlight reaches the ground in the cloud shadows or the gaps between them. To account for this properly in atmospheric models typically requires tracking the exact path of sunlight through the Earth-Atmosphere system, which can be a cumbersome task. We present an alternative approach to calculate how frequently different amounts of sunlight reach the surface under broken cloud scenes. This new approach uses just a few key pieces of information about the clouds to make predictions. Compared with existing approaches, the predicted surface sunlight is substantially more realistic. Results are consistent across two different algorithms tested and arrive at predictions in similar ways. The combined accuracy and efficiency of the new approach has several applications.

show abstract

“…The CNN is trained on a large database where the multi-spectral radiances are simulated with 3D RT and the pixel-scale cloud properties (typically, optical thickness and column-mean effective droplet radius) are known from the LES and assumed microphysics. Deep learning has thus been applied to both nadir satellite imagery [40] and to an array of ground-based sky cameras [41].…”

Section: Contextmentioning

confidence: 99%

Multi-View Polarimetric Scattering Cloud Tomography and Retrieval of Droplet Size

et al. 2020

View full text Add to dashboard Cite

Tomography aims to recover a three-dimensional (3D) density map of a medium or an object. In medical imaging, it is extensively used for diagnostics via X-ray computed tomography (CT). We define and derive a tomography of cloud droplet distributions via passive remote sensing. We use multi-view polarimetric images to fit a 3D polarized radiative transfer (RT) forward model. Our motivation is 3D volumetric probing of vertically-developed convectively-driven clouds that are ill-served by current methods in operational passive remote sensing. Current techniques are based on strictly 1D RT modeling and applied to a single cloudy pixel, where cloud geometry defaults to that of a plane-parallel slab. Incident unpolarized sunlight, once scattered by cloud-droplets, changes its polarization state according to droplet size. Therefore, polarimetric measurements in the rainbow and glory angular regions can be used to infer the droplet size distribution. This work defines and derives a framework for a full 3D tomography of cloud droplets for both their mass concentration in space and their distribution across a range of sizes. This 3D retrieval of key microphysical properties is made tractable by our novel approach that involves a restructuring and differentiation of an open-source polarized 3D RT code to accommodate a special two-step optimization technique. Physically-realistic synthetic clouds are used to demonstrate the methodology with rigorous uncertainty quantification.

show abstract

Retrieval of Cloud Optical Thickness from Sky-View Camera Images using a Deep Convolutional Neural Network based on Three-Dimensional Radiative Transfer

Cited by 26 publications

References 46 publications

An Intensive Campaign-Based Intercomparison of Cloud Optical Depth from Ground and Satellite Instruments under Overcast Conditions

An Intensive Campaign-Based Intercomparison of Cloud Optical Depth from Ground and Satellite Instruments under Overcast Conditions

On the Relationship Between Shallow Cumulus Cloud Field Properties and Surface Solar Irradiance

Multi-View Polarimetric Scattering Cloud Tomography and Retrieval of Droplet Size

Contact Info

Product

Resources

About